**Background.** As US states increasingly legalize medical and adult cannabis use, the diversity of cannabis products, consumption modes, and THC/CBD potencies has expanded dramatically. While smoking dried flower remains common, vaping flower and oils is growing, and concentrated products can approach 100% THC, elevating risks for psychosis, neurotoxicity, and cardiotoxicity. However, major nationally representative US surveillance surveys—including the National Survey on Drug Use and Health (NSDUH), National Health and Nutrition Examination Survey, and the Behavioral Risk Factor Surveillance System (BRFSS)—either omit product type and mode details entirely or capture only a single 'primary mode of use,' which may introduce substantial misclassification bias. This study uses detailed session-level data from the Releaf App to quantify the degree of clinically relevant misclassification that may occur when surveys omit cannabis product characteristics.
**Methods.** Data were drawn from the Releaf App, a free mobile application designed for tracking real-time subjective effects of cannabis use across products and consumption modes. The analysis was restricted to 2018 to avoid potential confounding from the 2019 EVALI outbreak and the COVID-19 pandemic. After excluding users under age 18, those outside the US, and sessions with incomplete data, the final analytic sample comprised 3,258 users who recorded 26,322 cannabis use sessions using 9,693 unique products. Users reported product type (flower, concentrate, edible, tincture, topical, pill, or 'other'), and for flower and concentrate sessions, inhalation method (joint, pipe, or vape). Optional fields included product potency (THC and CBD percentages) and user demographics (age, gender, state). Edibles and tinctures were assumed to be consumed orally; topicals and 'other' were collapsed as 'other' modes. Binary indicators were created for each product type, mode, and product-by-mode combination. Data were collapsed to the user level, and means with 95% confidence intervals were calculated. Analyses were conducted using Stata 15.1.
**Key Results.** The average user reported 28 sessions (median: 4) and 11.1 distinct products (median: 4). Among those reporting demographics, 56.9% were female, most were ages 25–44, and 38.4% lived in the West. Smoking was the most common primary mode (47.1%; 95% CI: 45.3%–48.8%), followed by vaping (36.5%; 95% CI: 34.9%–38.2%), eating/drinking (10.4%; 95% CI: 9.4%–11.5%), and other (1.4%; 95% CI: 1.1%–1.9%). However, 22.7% of users reported multimodal use: 17.1% (95% CI: 15.8%–18.4%) used two modes and 5.6% (95% CI: 4.8%–6.4%) used three or more modes over 12 months. Any smoking was reported by 56.3% (95% CI: 54.6%–58.0%), any vaping by 47.3% (95% CI: 45.6%–49.0%), any edible by 12.0% (95% CI: 10.9%–13.2%), any tincture by 10.5% (95% CI: 9.5%–11.6%), and any other product by 3.7% (95% CI: 3.1%–4.5%). Product type varied substantially within mode: among vapers, 58.5% vaped concentrates only, 31.3% vaped flower only, and 10.2% vaped both. Among smokers, 91.8% smoked flower only, 4.6% smoked concentrates only, and 3.5% smoked both. THC potency was 3.4 times higher for concentrates (60.5%; 95% CI: 58.5%–62.5%) than for flower (17.9%; 95% CI: 17.4%–18.4%). CBD potency was 3.1 times higher for concentrates (21.4%; 95% CI: 19.3%–23.4%) than for flower (7.0%; 95% CI: 6.5%–7.5%).
**Clinical Implications.** These findings demonstrate that relying on 'primary mode of use' or any recent cannabis use without product details—as is standard in current US surveillance surveys—conceals substantial heterogeneity with direct clinical relevance. The 3.4-fold difference in THC potency between flower and concentrates has implications for psychosis risk, neurotoxicity, and cardiotoxicity. The finding that nearly one-third of vapers use only flower (not concentrates) means that assuming vaping equals concentrate use would misclassify a large proportion of users. The study's limitations include non-representative sampling (overrepresentation of medical users, females, and Western residents), potential underreporting of sessions, and reliance on self-reported potency. Nevertheless, the consistency of self-reported THC potencies with Washington state dispensary data supports the validity of these measures. The authors conclude that nationally representative surveillance surveys must urgently incorporate detailed questions on product type, consumption mode, and potency to inform clinical decision-making, cannabis policy, product safety standards, and interventions to reduce high-risk use.